Research
NewOpen-Sourcing Post-Training Pipeline for Xiaomi-Robotics-0
Xiaomi-Robotics-0 achieves precise earbud-to-case insertion using just 20 hours of post-training data. We open-source the full post-training pipeline.
Xiaomi-Robotics-0: An Open-Sourced Vision-Language-Action Model with Real-Time Execution
Xiaomi-Robotics-0 is an advanced Vision-Language-Action (VLA) model optimized for high performance and real-time execution.
Our Mission
We are building general robots with world-leading Embodied Foundation Models. True intelligence must go beyond digital screens to understand and change the real world. By unifying digital reasoning with physical action, we are shattering the boundary between bits and atoms, endowing robots with a true embodied brain.
As pioneers at the absolute frontier of Physical AI, we are guided by long-termism and an uncompromising pursuit of innovation. We are dedicated to building end-to-end foundation models that enable machines to seamlessly perceive, reason, and act.
We are seeking visionaries who combine top-tier academic vision with the engineering firepower to scale these massive models into reality. Join us in the uncharted to define the intelligent future of the physical world.